Striping has no redundancy at all: every file is spread across every member, so losing one disk means losing a slice of nearly everything. Recoverable — but the odds are set by the condition of the disk that failed.
RAID 0 keeps no redundant copy, so recovery means reading every member — including the failed one — and reassembling the stripe from images. How much returns depends on how much of the casualty can be read.
RAID 0 splits data into blocks and writes them across the members in rotation, which is what makes it fast and what makes it fragile. There is no parity and no mirror, so nothing can be computed back. Lose one disk and you lose one block in every few — enough that almost no file survives whole, which is why the volume stops mounting rather than degrading gracefully. Two disks in RAID 0 are roughly twice as likely to fail as one, and the consequence of a failure is total rather than partial.
Every member is imaged read-only, including the failed one after any physical repair it needs — donor heads, a board repair, or firmware work to get it enumerating long enough to read. The stripe parameters are then derived from the data itself: block size, disk order and start offset, established by testing candidates until the reassembly produces coherent file-system structures rather than noise. The volume is assembled virtually from the images and the files extracted from there.
If the casualty images completely, recovery is generally full. If parts of it are unreadable, the missing blocks land scattered through the volume, and the effect varies by file type: large files are more likely to be touched, and formats that tolerate gaps — some video and image containers — may still open with defects, while databases and archives usually will not. We report which files fall in an affected region by name rather than returning silently damaged copies. That distinction is the difference between a recovery and a file count.
Power the machine down rather than retrying it. Don’t let a controller or operating system re-create or initialise the array to “repair” it — that writes new metadata over the structures needed to work out the original geometry. Label each disk with its bay position before removing it, since disk order is central to the reassembly and guessing it costs time. If any disk is clicking or grinding, leave it powered off.
Usually yes, though RAID 0 offers no redundancy, so the outcome tracks the condition of the failed disk. Every member is imaged read-only, the failed one is repaired physically first if it needs it, and the stripe is reassembled from those images. A disk that images fully generally means a full recovery.
Because striping writes each file in blocks spread across all the members in turn. Losing one disk removes a block from nearly every file, so the file system can no longer be read as a whole. That is the trade-off RAID 0 makes: speed and capacity in exchange for no fault tolerance whatsoever.
All of them. Each disk holds a different part of every file, so the stripe cannot be reassembled without every member present. Label them by bay order before removing them — disk order matters for the reconstruction.
From £500 plus VAT for RAID, NAS and server work, fixed in a written quote after a free 48-hour diagnostic. Where the failed member needs physical drive-level work there is a 50% deposit, with the balance payable only on success.
No. Creating or initialising an array writes new metadata to the disks and can destroy the very structures a recovery would use to determine the original stripe geometry. If the data matters, stop at the point of discovery and send the disks in as they are.